CoChat
Sign Up
Cited by 1,778Face recognition and analysisEmotion and Mood RecognitionFace and Expression Recognition

Automatic analysis of facial expressions: the state of the art

Open access
Pantie, M., Rothkrantz, L.J.M.·2000·IEEE Transactions on Pattern Analysis and Machine Intelligence
DOI 10.1109/34.895976
Open this paper in CoChat — free accountOpen PDF

Abstract

Humans detect and interpret faces and facial expressions in a scene with little or no effort. Still, development of an automated system that accomplishes this task is rather difficult. There are several related problems: detection of an image segment as a face, extraction of the facial expression information, and classification of the expression (e.g., in emotion categories). A system that performs these operations accurately and in real time would form a big step in achieving a human-like interaction between man and machine. The paper surveys the past work in solving these problems. The capability of the human visual system with respect to these problems is discussed, too. It is meant to serve as an ultimate goal and a guide for determining recommendations for development of an automatic facial expression analyzer.

AI snapshot

Written from the paper’s full text.

TL;DR

This survey reviews the state of the art (through 2000) in automatic facial expression analysis, decomposing the problem into face detection, facial expression data extraction, and facial expression classification. It compares surveyed systems against an "ideal analyzer" modeled on the human visual system and concludes that none of the three subproblems has been solved in the general case.

Key findings

  • The paper identifies three related subproblems — face detection, facial expression information extraction, and facial expression classification — and argues an ideal system must automate all three, ideally in real time for HCI use.
  • Early systems (1991–1995) achieved correct recognition rates of 70–92 percent when recognizing 3–8 emotion categories; none performed quantified expression classification in terms of facial actions, and only Moses et al. [57] performed in real time.
  • Among the surveyed recent systems, reported recognition rates include 92 percent (Yoneyama et al., four expressions), 89 percent for familiar subjects and 73 percent for unknown persons (Hong et al.), 84.5 percent (Huang and Huang), 92 percent generalization / 75 percent for a novel subject (Lyons et al.), 90.1 percent (Zhang et al.), 92 percent upper-face AUs and 86 percent lower-face AUs (Pantic and Rothkrantz), 92/88/83 percent for eyebrow/eye/nose-mouth regions (Cohn et al.), 98 percent (Essa and Pentland), 95 percent (Wang et al.), and 88 percent (Black and Yacoob).
  • Most systems rely on frontal-view images of faces without facial hair or glasses, assume constant illumination and immobile subjects, and perform only singular classification into Ekman's six basic emotion categories; only two systems ([62], [106]) perform quantified classification into multiple emotion categories, and none quantifies facial action codes or distinguishes all 44 FACS AUs.

Methods

  • Design: a survey/comparative review of vision-based automatic facial expression analysis, covering work from the mid-1970s onward with emphasis on the late 1990s; systems are tabulated (Tables 2–8) against properties of an "ideal analyzer" (Table 1).
  • Scope: selectively discusses systems addressing face detection, facial expression data extraction (template-based vs. feature-based; holistic, analytic, and hybrid face representations), and classification (template-, neural-network-, and rule-based methods) from both static images and image sequences.
  • Analysis: qualitative comparison of reported techniques, assumptions, and recognition rates; the authors deliberately refrain from ranking systems, arguing that no single common testing database exists.

Limitations

  • Stated by the paper: no single well-defined, commonly used testing database exists, so the authors deliberately do not rank systems objectively; several surveyed methods were tested only on the images used for training, so their person-independence is uncertain; most systems assume frontal views, constant illumination, no facial hair or glasses, and immobile subjects, which is unrealistic for the target application domains; only two systems handle automatic face detection in an arbitrary scene and only three deal to some extent with rigid head motions; only Essa and Pentland's method handles faces with facial hair and/or eyeglasses; no system performs quantification of facial action codes or distinguishes all 44 FACS AUs.
  • Evident from the text: the survey is selective rather than exhaustive, so coverage of the field is incomplete by the authors' own statement; the "ideal analyzer" is a normative construct rather than an implemented benchmark, and the comparison is largely qualitative with heterogeneous, non-comparable evaluation protocols across systems.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1109/34.895976

Cites

107 of 107 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 124 cited sources checked · 0 supported · 0 flagged

  • Findingcitation

    The majority of studies treat various human communication channels separately.

    “So far, the majority of studies treat various human communication channels separately, as indicated by Nakatsu [58].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Nakatsu [58]

      Not checked in the free run.

  • Findingcitation

    Relatively few existing works combine different modalities into a single system for human communicative reaction analysis.

    “Relatively few existing works combine different modalities into a single system for human communicative reaction analysis.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [9] source

      Not checked in the free run.

    • cites [15]

      Not checked in the free run.

  • Findingnone given

    Most studies on automated expression analysis perform an emotional classification.

    “Most of the studies on automated expression analysis perform an emotional classification.”
    Quote in the paperNot a substantial claim
  • Findingown data

    Several surveyed systems can be classified as methods for facial expression analysis from static images.

    “As shown in Table 3, several surveyed systems can be classified as methods for facial expression analysis from static images.”
    Quote in the paperNot a substantial claim
  • Findingnone given

    None of the systems in Table 2 performs a quantified expression classification in terms of facial actions.

    “None of these systems performs a quantified expression classification in terms of facial actions.”
    Quote in the paperNot a substantial claim
  • Classified as probably not a substantial claim. 58
  • May combine two claims. 47
  • The quote only partly states this claim. 6

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingcitation

    Although Hong et al. utilize the PersonSpotter system, which deals with real-time processing of video sequences, they perform facial expression analysis from static images.

    “Although Hong et al. utilize the PersonSpotter system [81], which deals with real-time processing of video sequences, they perform facial expression analysis from static images.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [81] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Pantic and Rothkrantz's face detection system does not allow facial hair or glasses.

    “No facial hair or glasses are allowed.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [62] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingcitation

    Kobayashi and Hara's face detection system requires the observed subject to face the camera while sitting at approximately 1m distance in front of it.

    “The observed subject should face the camera while siting at approximately 1m distance in front of it.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [42]

      Not checked in the free run.

  • May combine two claims.

    Findingnone given

    None of the surveyed systems are able to perform facial expression information extraction from images of occluded faces or a quantified expression classification in terms of facial actions.

    “None of these systems are able to perform facial expression information extraction from images of occluded faces or a quantified expression classification in terms of facial actions.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingnone given

    Systems for automatic analysis and synthesis of facial expressions primarily concern facial expression animation and do not attempt to classify the observed facial expression either in terms of facial actions or in terms of emotion categories.

    “These systems primarily concern facial expression animation and do not attempt to classify the observed facial expression either in terms of facial actions or in terms of emotion categories.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingcitation

    The AAM search algorithm has been tested on 100 hand labeled face images, and 19.2 percent failed to converge to a satisfactory result.

    “The AAM search algorithm has been tested on 100 hand labeled face images.Of these, 19.2 percent failed to converge to a satisfactory result [12].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [12] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given

    Since the mid 1970s, different approaches have been proposed for facial expression analysis from either static facial images or image sequences.

    “Since the mid 1970s, different approaches are proposed for facial expression analysis from either static facial images or image sequences.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingcitation

    The method of Pentland et al. is real-time and has been successfully tested on a database of 7,562 images of some 3,000 people of both sexes, ranged in age and ethnicity, having varying head positions, headwear, and facial hair.

    “The method of Pentland et al. [65] is real-time and has been successfully tested on a database of 7,562 images of some 3,000 people of both sexes, ranged in age and ethnicity, having varying head positions, headwear, and facial hair.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [65]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingreasoning

    Depending on the face model, a template-based or a feature-based method is applied for facial expression data extraction.

    “Depending on the face model a template-based or a feature-based method is applied for facial expression data extraction.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingcitation

    The utilized techniques include holistic spatial analysis, spatio-temporal analysis, and analytic spatial analysis.

    “The utilized techniques include holistic spatial analysis [13], [70], [54], spatio-temporal analysis [52], [100], [73], [57], and analytic spatial analysis [93], [43], [44], [92], [40].”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [13] source

      Not checked in the free run.

    • cites [70] source

      Not checked in the free run.

    • cites [54] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingcitation

    Edwards et al.'s method works with images of faces without facial hair and glasses, which are hand-labeled with the landmark points beforehand approximated with the proposed AAM.

    “The method works with images of faces without facial hair and glasses, which are hand-labeled with the landmark points beforehand approximated with the proposed AAM.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [18] source

      Not checked in the free run.

  • May combine two claims.

    Findingcitation

    Some methods listed in Table 2 do not perform automatic facial data extraction; others achieve this by using facial motion analysis.

    “Some of the methods listed in Table 2 do not perform automatic facial data extraction (see column 8); the others achieve this by using facial motion analysis [52], [100], [101], [73], [57].”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [52]

      Not checked in the free run.

    • cites [100]

      Not checked in the free run.

    • cites [101] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Detection of the exact face position in an observed image or image sequence has been approached in two ways: holistic and analytic.

    “Independently of the kind of input imagesÐfacial images or arbitrary imagesÐdetection of the exact face position in an observed image or image sequence has been approached in two ways.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Two of the works surveyed in this paper perform automatic face detection in an arbitrary scene.

    “Two of the works surveyed in this paper perform automatic face detection in an arbitrary scene.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingcitation

    Huang and Huang's face detection system requires the face to be without facial hair and glasses, no rigid head motion, and linear illumination variations to work correctly.

    “The face should be without facial hair and glasses, no rigid head motion may be encountered and illumination variations must be linear for the system to work correctly.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [32]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingcitation

    The PersonSpotter system performs well in the presence of background motion but fails in the case of covered or too much rotated faces.

    “Steffens et al. [81] reported that their system performs well in the presence of background motion, but fails in the case of covered or too much rotated faces.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [81] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingcitation

    Yoneyama et al. do not state which method has been applied and do not report a limitation of the used method.

    “It is not stated which method has been applied and no limitation of the used method has been reported by Yoneyama et al.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [104]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given

    There is no psychological scrutiny on classification of facial expressions into multiple emotion categories.

    “Still, there is no psychological scrutiny on this topic.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingcitation

    In most work in automatic facial expression analysis, the camera is either mounted on a helmetlike device worn by the subject or placed so that the image has the face in frontal view.

    “The camera is either mounted on a helmetlike device worn by the subject (e.g., [62], [59]) or placed in such a way that the image has the face in frontal view.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [62] source

      Not checked in the free run.

    • cites [59] source

      Not checked in the free run.

  • May combine two claims.

    Findingnone given

    A first category of surveyed static-image systems utilizes a holistic or a hybrid approach to face representation and applies a template-based method for facial expression information extraction from an input image.

    “A first category of these utilizes a holistic or a hybrid approach to face representation (Table 5) and applies a template-based method for facial expression information extraction from an input image.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingcitation

    Kimura and Yachida's face detection system requires the face to be without facial hair and glasses and in a direct face-to-face position with the camera.

    “The face should be without facial hair and glasses and in a direct face-to-face position with the camera [99].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [99] source

      Not checked in the free run.

  • May combine two claims.

    Findingnone given

    The surveyed facial expression analyzers classify the encountered expression either as a particular facial action or a particular basic emotion, and some systems perform both.

    “The surveyed facial expression analyzers classify the encountered expression (i.e., the extracted facial expression information) either as a particular facial action or a particular basic emotion.Some of the systems perform both.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingcitation

    Pantic and Rothkrantz's system cannot deal with minor inaccuracies of the extracted facial data and deals merely with images of faces without facial hair or glasses.

    “The system cannot deal with minor inaccuracies of the extracted facial data and it deals merely with images of faces without facial hair or glasses.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [62] source

      Not checked in the free run.

  • May combine two claims.

    Findingnone given

    A first category of the surveyed methods for automatic expression analysis from static images applies a template-based method for expression classification, and the methods in this category perform expression classification into a single basic emotion category.

    “A first category of the surveyed methods for automatic expression analysis from static images applies a template-based method for expression classification.The methods in this category perform expression classification into a single basic emotion category.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingcitation

    In Black and Yacoob's approach, the initial regions for the head and the facial features were selected by hand and thereafter automatically tracked.

    “In the approach proposed by Black and Yacoob, the initial regions for the head and the facial features were selected by hand and thereafter automatically tracked.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [2] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingcitation

    Zhang et al. deal only with frontal view images of nine female Japanese subjects, manually normalized so that the distance between the eyes is 60 pixels.

    “Zhang et al. deal only with PST Â PST pixels frontal view images of nine female Japanese subjects, manually normalized so that the distance between the eyes is 60 pixels.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [106]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Essa and Pentland's method has been applied to frontal-view facial image sequences.

    “The method has been applied to frontal-view facial image sequences.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [24] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Lyons et al. used a similar face representation for expression classification into the six basic plus neutral emotion categories.

    “A similar face representation was recently used by Lyons et al. [51] for expression classification into the six basic plus ªneutralº emotion categories.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [51] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Huang and Huang's PDM was generated from 90 facial feature points manually localized in 90 images of 15 Chinese subjects showing six basic emotions.

    “The used PDM has been generated from 90 facial feature points that have been manually localized in 90 images of 15 Chinese subjects showing six basic emotions (Figs. 3 and 4).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [32]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given

    Independent of the used classification categories, the mechanism of classification applied by a particular surveyed expression analyzer is either a template-based, a neural-network-based, or a rule-based classification method.

    “Independent of the used classification categories, the mechanism of classification applied by a particular surveyed expression analyzer is either a template-based-or a neural-network-based-or a rule-based-classification method.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Hong et al. have not discussed the issue of using the dense model-graph for facial action coding based on extracted deformations of the graph.

    “However, this issue has not been discussed by Hong et al. [30].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [30] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Kimura and Yachida have not discussed the issue of using the Potential Net for facial action encoding based on extracted deformations of the net.

    “However, Kimura and Yachida have not discussed this issue.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [41]

      Not checked in the free run.

  • May combine two claims.

    Findingnone given

    A first category of surveyed approaches to automatic facial expression analysis from image sequences uses a holistic or a hybrid approach to face representation and applies a template-based method for facial expression information extraction from an input image sequence.

    “A first category of the surveyed approaches to automatic facial expression analysis from image sequences uses a holistic or a hybrid approach to face representation (Table 3, Table 5) and applies a template-based method for facial expression information extraction from an input image sequence.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingnone given

    The second category of surveyed methods for automatic facial expression analysis from static images uses an analytic approach to face representation and applies a feature-based method for expression information extraction from an input image.

    “The second category of the surveyed methods for automatic facial expression analysis from static images uses an analytic approach to face representation (Table 3, Table 5) and applies a feature-based method for expression information extraction from an input image.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingcitation

    Zhao et al. utilize a point-based frontal-view face model but do not deal with automatic facial expression data extraction.

    “Zhao et al. [107] also utilize a point-based frontal-view face model but do not deal with automatic facial expression data extraction.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [107] source

      Not checked in the free run.

  • May combine two claims.

    Findingcitation

    Yoneyama et al.'s method requires the face to be without facial hair and glasses and no rigid head motion to work correctly.

    “The face should be without facial hair and glasses and no rigid head motion may be encountered for the method to work correctly.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [104]

      Not checked in the free run.

  • May combine two claims.

    Findingcitation

    Padgett and Cottrell use a holistic face representation but do not deal with facial expression information extraction in an automatic way.

    “Padgett and Cottrell [61] also use a holistic face representation, but they do not deal with facial expression information extraction in an automatic way.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [61] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Wang et al.'s method assumes the initial location of the FFPs in the first frame of an input image sequence is known.

    “The initial location of the FFPs in the first frame of an input image sequence is assumed to be known.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [95] source

      Not checked in the free run.

  • May combine two claims.

    Findingcitation

    The performance of Pantic and Rothkrantz's detection scheme was tested on 496 dual views, and human observers in 89 percent approved when visually inspected the achieved localization of the facial features.

    “The performance of the detection scheme was tested on 496 dual views.Human observers in 89 percent approved when visually inspected the achieved localization of the facial features.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [62] source

      Not checked in the free run.

  • May combine two claims.

    Findingnone given

    A second category of the surveyed methods for automatic facial expression analysis from static images applies a neural network for facial expression classification, and except the method proposed by Zhang et al., the methods in this category perform facial expression classification into a single basic emotion category.

    “A second category of the surveyed methods for automatic facial expression analysis from static images applies a neural network for facial expression classification.Except the method proposed by Zhang et al. [106], the methods belonging to this category perform facial expression classification into a single basic emotion category.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingcitation

    The real-time system developed by Kobayashi and Hara works with online taken images of subjects with no facial hair or glasses facing the camera while sitting at approximately 1m distance from it.

    “The real-time system developed by Kobayashi and Hara works with online taken images of subjects with no facial hair or glasses facing the camera while siting at approximately 1m distance from it.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [42]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    The first category of the surveyed methods for automatic facial expression analysis from facial image sequences applies a template-based method for expression classification.

    “The first category of the surveyed methods for automatic facial expression analysis from facial image sequences applies a template-based method for expression classification.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingreasoning

    The Black and Yacoob method does not deal with blends of emotional expressions; for example, a blend of the angry and scared expressions is recognized as disgust.

    “Also, the method does not deal with blends of emotional expressions.For example, a blend of the angry and scared expressions is recognized as disgust.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [2] source

      Not checked in the free run.

    • cites [3]

      Not checked in the free run.

  • May combine two claims.

    Findingnone given

    Common assumptions in surveyed systems include frontal facial view, constant illumination, fixed light source, absence of facial hair or glasses, young subjects, and subjects of the same ethnicity.

    “Some common assumptions are: the images contain frontal facial view, the illumination is constant, the light source is fixed, the face has no facial hair or glasses, the subjects are young (i.e., without permanent wrinkles), and of the same ethnicity.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    In a message, the verbal part contributes 7 percent, the vocal part 38 percent, and facial expression 55 percent to the effect of the message as a whole.

    “Mehrabian indicated that the verbal part (i.e., spoken words) of a message contributes only for 7 percent to the effect of the message as a whole, the vocal part (e.g., voice intonation) contributes for 38 percent, while facial expression of the speaker contributes for 55 percent to the effect of the spoken message [55].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [55] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Human encoding of the visual stimulus (face and its expression) may be in the form of a primal sketch and may be hardwired.

    “As indicated by Ellis [23], human encoding of the visual stimulus (face and its expression) may be in the form of a primal sketch and may be hardwired.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Ellis [23]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    FACS provides a linguistic description of all possible, visually detectable, facial changes in terms of 44 Action Units.

    “In a form of rules, FACS provides a linguistic description of all possible, visually detectable, facial changes in terms of 44 so-called Action Units (AUs).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [21]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Premisereasoning

    The face plays an essential role in interpersonal communication.

    “The terms ªface-to-faceº and ªinterfaceº indicate that the face plays an essential role in interpersonal communication.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Two-gray-level images of 100 to 200 pixels form a lower limit for detection of a face by a human observer.

    “It is generally believed that two-gray-levels images of 100 to 200 pixels form a lower limit for detection of a face by a human observer [75], [8].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [75] source

      Not checked in the free run.

    • cites [8]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Ekman defined six basic emotion categories: happiness, sadness, surprise, fear, anger, and disgust.

    “Ekman defined six such categories, referred to as the basic emotions: happiness, sadness, surprise, fear, anger, and disgust [19].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [19] source

      Not checked in the free run.

  • May combine two claims.

    Premisenone given

    Human mechanisms for face detection are very robust, but interpretation of facial expressions is not.

    “While the human mechanisms for face detection are very robust, the same is not the case for interpretation of facial expressions.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Premisenone given

    In practice, it may not be possible to define a face model that can both reflect each and every change in facial appearance and have features detectable in a facial image or image sequence.

    “In practice, it may not be possible to define a face model that can satisfy both, to reflect each and every change in facial appearance and whose features are detectable in a facial image or image sequence.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The appearance of upper face features plays a more important role in face interpretation than lower face features.

    “It is interesting to note that the appearance of the upper face features plays a more important role in face interpretation as opposed to lower face features [22].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [22] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Recently, the problem of automatic face detection in an arbitrary scene has drawn great attention.

    “Recently, the problem of automatic face detection in an arbitrary scene has drawn great attention (e.g., see [74], [83], [85]).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [74] source

      Not checked in the free run.

    • cites [83] source

      Not checked in the free run.

    • cites [85] source

      Not checked in the free run.

    • Rests on an assumption: In most real-life situations where an automated facial expression analyzer is to be employed, the location of a face in the image is not known a priori.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given

    Information about the context in which a facial expression appears is very difficult to obtain in an automatic way.

    “Nevertheless, the information about the context in which a facial expression appears is very difficult to obtain in an automatic way.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The approaches explored lately include systems for automatic analysis and synthesis of facial expressions.

    “The approaches that have been explored lately also include systems for automatic analysis and synthesis of facial expressions [86], [56], [39], [47], [89], [88], [53], [14], [19]”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [86] source

      Not checked in the free run.

    • cites [56] source

      Not checked in the free run.

    • cites [39] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Ekman described each basic emotion in terms of a facial expression that uniquely characterizes that emotion.

    “He described each basic emotion in terms of a facial expression that uniquely characterizes that emotion.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [19] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given

    Ekman's description of the six prototypic facial expressions of emotion is linguistic and thus ambiguous.

    “First, Ekman's description of the six prototypic facial expressions of emotion is linguistic (and, thus, ambiguous).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisenone given

    It is not certain that each facial expression able to be displayed on the face can be classified under the six basic emotion categories.

    “Also, it is not at all certain that each facial expression able to be displayed on the face can be classified under the six basic emotion categories.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Premisenone given

    It is often very difficult to determine the exact nature of the expression on a person's face.

    “It is often very difficult to determine the exact nature of the expression on a person's face.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    A trained observer can correctly classify faces showing six basic emotions with an average of 87 percent.

    “According to Bassili [1], a trained observer can correctly classify faces showing six basic emotions with an average of 87 percent.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Bassili [1]

      Not checked in the free run.

  • May combine two claims.

    Study design and conductnone given

    The method of Cohn et al. requires the face to be without facial hair or glasses, no rigid head motion, an expressionless first frame, and facial landmark points marked in the first frame for the method to work correctly.

    “The face should be without facial hair/ glasses, no rigid head motion may be encountered, the first frame should be an expressionless face, and the facial landmark points should be marked in the first frame for the method to work correctly.”
    Quote in the paperNot a substantial claim

  • The quote only partly states this claim.

    Premisecitation

    The Lyons et al. method achieved a generalization rate of 75 percent for recognition of expression of a novel subject when the image set was partitioned into nine segments, each corresponding to one expresser.

    “The generalization rate was 75 percent for recognition of expression of a novel subject.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [51] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The time necessary for performing a full analysis of an incoming facial image by the Hong et al. method is about eight seconds.

    “The time necessary for performing a full analysis of an incoming facial image is about eight seconds.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [30] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    There are infinitely many combinations of different facial actions and their intensities that should be modeled with a finite set of templates.

    “There are infinitely a lot of combinations of different facial actions and their intensities that should be modeled with a finite set of templates.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given

    It is not known how the Edwards et al. method will behave in the case of an unknown subject.

    “It is not known how the method will behave in the case of an unknown subject.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    It is not known how the Huang and Huang method will behave in the case of unknown subjects.

    “It is not known how the method will behave in the case of unknown subjects.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given

    The problem of template-based recognition becomes even more difficult because everybody has his or her own maximal intensity of displaying a certain facial action.

    “The problem becomes even more difficult due to the fact that everybody has his/her own maximal intensity of displaying a certain facial action.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The Hara and Kobayashi neural network achieved an average recognition rate of 85 percent.

    “The average recognition rate was 85 percent.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [42]

      Not checked in the free run.

  • May combine two claims. The quote only partly states this claim.

    Premisecitation

    The Lyons et al. method achieved a generalization rate of 92 percent when tested on 193 images divided into 10 segments.

    “The generalization rate was 92 percent.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [51] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Interpretationnone given

    The dense model-graph seems very suitable for facial action coding based on the extracted deformations of the graph.

    “The dense model-graph seems very suitable for facial action coding based on the extracted deformations of the graph.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [30] source

      Not checked in the free run.

  • May combine two claims.

    Interpretationnone given

    The descriptions of the emotional expressions given in terms of facial actions are incomplete; for example, an expression with lowered mouth corners and raised eyebrows will be classified as sadness.

    “Also, the descriptions of the emotional expressions, given in terms of facial actions, are incomplete.For example, an expression with lowered mouth corners and raised eyebrows will be classified as sadness.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    The Hong et al. method achieved a recognition rate of 89 percent in the case of familiar subjects and 73 percent in the case of unknown persons.

    “The achieved recognition rate was 89 percent in the case of the familiar subjects and 73 percent in the case of unknown persons.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [30] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The Yoneyama et al. method achieved an average recognition rate of 92 percent.

    “The average recognition rate was 92 percent.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [104]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductcitation

    The images used for training of the Yoneyama et al. networks were also used for their testing.

    “The images used for training of the networks were also used for their testing.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [104]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    The Potential Net method seems suitable for facial action encoding based on the extracted deformations of the net.

    “The method seems suitable for facial action encoding based on the extracted deformations of the net.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [41]

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    The expression recognition performance of the AAM has been trained and tested on P Â PHH images of six basic emotional expressions shown by 25 subjects, and the achieved recognition rate for the six basic and neutral emotion categories was 74 percent.

    “The expression recognition performance of the AAM has been trained and tested on P Â PHH images of six basic emotional expressions shown by 25 subjects.The images were chosen for limited pose and lightning variation.The achieved recognition rate for the six basic and ªneutralº emotion categories was 74 percent.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [18] source

      Not checked in the free run.

  • May combine two claims.

    Interpretationnone given

    A typical neural network can perform a quantified facial expression categorization into multiple classes while, in general, template-based methods cannot achieve such a performance.

    “We are doing so because a typical neural network can perform a quantified facial expression categorization into multiple classes while, in general, the template-based methods cannot achieve such a performance.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    The Pantic and Rothkrantz system achieved an average recognition rate of 92 percent for the upper face AUs and 86 percent for the lower face AUs.

    “The average recognition rate was 92 percent for the upper face AUs and 86 percent for the lower face AUs.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [62] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given

    It is not known how the Zhao and Kearney method will behave in the case of an unknown subject.

    “It is not known how the method will behave in the case of an unknown subject.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationnone given

    Black and Yacoob did not give a full list of the midlevel predicates and the number of different facial actions that the method can recognize is not known.

    “Black and Yacoob did not give a full list of the midlevel predicates and the number of different facial actions that the method can recognize is not known.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

    Premisecitation

    The Wang et al. method had an average process time of 2.5s/frame.

    “The average process time was 2.5s/frame.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [95] source

      Not checked in the free run.

  • The quote only partly states this claim.

    Premisecitation

    The Wang et al. method achieved an average recognition rate of 95 percent.

    “The average recognition rate was 95 percent.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [95] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductcitation

    The dual-views used for testing of the Pantic and Rothkrantz system have been recorded under constant illumination and none of the subjects had a moustache, a beard, or wear glasses.

    “The dual-views used for testing of the system have been recorded under constant illumination and none of the subjects had a moustache, a beard, or wear glasses.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [62] source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Interpretationnone given

    The problem of facial expression analysis has three different but related aspects: face detection, facial expression information extraction, and facial expression classification.

    “We have identified three different but related aspects of the problem: face detection, facial expression information extraction, and facial expression classification.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    A very small number of training examples (six sequences) and an insufficient diversity of the subjects (one person) have probably caused the unsuccessful testing of the Kimura and Yachida method.

    “A very small number of training examples (six sequences) and an insufficient diversity of the subjects (one person) have probably caused this.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S1 (asserted), S3 (load-bearing). May combine two claims.

    Premisecitation

    The human visual system assigns a higher priority to upper face features than to lower face features because they play a more important role in facial expression interpretation.

    “Another appealing but still not investigated property of the human visual system is assigning a higher ªpriorityº to the upper face features than to the lower face features since they play a more important role in facial expression interpretation [22].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [22] source

      Not checked in the free run.

  • The quote only partly states this claim.

    Premisecitation

    The overall performance of the Pantic and Rothkrantz system achieved a correct recognition ratio of 91 percent on a set of 265 dual facial views representing six basic and various blended emotional expressions shown by eight subjects.

    “A correct recognition ratio of 91 percent has been reported.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [62] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductcitation

    The Otsuka and Ohya HMM was trained on 120 image sequences shown by two male subjects and tested on image sequences shown by the same subjects.

    “The HMM was trained on 120 image sequences, shown by two male subjects.The method was tested on image sequences shown by the same subjects.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [60] source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    The Cohn et al. method achieved classification accuracy of 92 percent for the eyebrow region, 88 percent for the eye region and 83 percent for the nose/mouth region.

    “The accuracy of the classification was 92 percent for the eyebrow region, 88 percent for the eye region and 83 percent for the nose/ mouth region.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [10] source

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    The method proposed by Cohn et al. deals neither with image sequences containing several facial actions in a row, nor with inaccurate facial data, nor with facial action intensity.

    “The method proposed by Cohn et al. deals neither with image sequences containing several facial actions in a row, nor with inaccurate facial data, nor with facial action intensity (yet the concept of the method makes it possible).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [10] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductcitation

    The Zhao and Kearney neural network was trained and tested on the whole set of data (94 images).

    “The NN was trained and tested on the whole set of data (94 images) with 100 percent recognition rate.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [107] source

      Not checked in the free run.

  • The quote only partly states this claim.

    Premisecitation

    The Black and Yacoob method achieved a recognition rate of 88 percent when tested on 70 image sequences containing 145 expressions shown by 40 subjects.

    “The achieved recognition rate was 88 percent.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [2] source

      Not checked in the free run.

    • cites [3]

      Not checked in the free run.

93 of 93

  • The paper relies on this without support.

    A face is perceived as a whole, not as a collection of the facial features.

    “Another characteristic of the human visual system is that a face is perceived as a whole, not as a collection of the facial features.”

  • The paper relies on this without support.

    When a face is partially occluded, humans perceive a whole face as if the perceptual system fills in the missing parts.

    “When a face is partially occluded (e.g., by a hand), we perceive a whole face, as if our perceptual system fills in the missing parts.”

  • The paper relies on this without support.

    Each person has his/her own maximal intensity of displaying a particular facial expression.

    “On the other hand, each person has his/her own maximal intensity of displaying a particular facial expression.”

  • The paper relies on this without support.

    In most real-life situations, complete immovability of the observed subject cannot be assumed.

    “In most real-life situations, complete immovability of the observed subject cannot be assumed.”

  • The paper relies on this without support.

    Considering the state-of-the-art in image processing techniques, inaccurate, noisy, and missing data could be expected.

    “Considering the stateof-the-art in image processing techniques, inaccurate, noisy, and missing data could be expected.”

  • The paper relies on this without support.

    Long delays make interaction desynchronized and less efficient.

    “Long delays make the interaction desynchronized and less efficient.”

  • The paper relies on this without support.

    In most real-life situations where an automated facial expression analyzer is to be employed, the location of a face in the image is not known a priori.

    “Yet, in most of the real-life situations where an automated facial expression analyzer is to be employed (e.g., in a multimodal/media HCI), the location of a face in the image is not known a priori.”

  • The paper relies on this without support.

    In most real-life situations, the observed subject will not remain immovable.

    “Also, in most of the real-life situations, it cannot be assumed that the observed subject will remain immovable, as assumed by some methods.”

  • The paper relies on this without support.

    Pure emotional expressions are seldom elicited.

    “First, ªpureº emotional expressions are seldom elicited.”

  • The paper relies on this without support.

    Most of the time, people show blends of emotional expressions.

    “Most of the time, people show blends of emotional expressions.”

  • The paper relies on this without support.

    Some facial actions such as blink, wink, and lips sucked into the mouth are either encountered or not.

    “Second, some facial actions such as blink, wink, and lips sucked into the mouth, are either encountered or not.”

  • The paper relies on this without support.

    Each person has his or her own maximal intensity of displaying a particular facial action.

    “In addition, each person has his/her own maximal intensity of displaying a particular facial action.”

  • The paper relies on this without support.

    FACS provides an option to score intensity on a 3-level intensity scale for only five different AUs.

    “First, there are merely five different AUs for which FACS provides an option to score intensity on a 3-level intensity scale (low, medium, and high).”
13 of 13

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Conclusion says all systems classify only basic emotions

    Major

    The conclusion says every surveyed approach performs only basic-emotion classification, but the discussion states that four surveyed systems also perform facial-action coding.

  • Citation for Ekman's six categories points to different authors

    Minor

    The statement attributes the six categories to Ekman and cites reference [19], but the paper's reference [19] is by Eisert and Girod.

2 of 2

Cites

Flagged first, then by citations

4 of the 111 references couldn’t be matched to a known paper.

Research works printed but not linked: 13

  1. wiskott, 1995, labelled graphs and dynamic link matching for face recognition and scene analysis, Reihe Physik, 53
  2. cottrell, 1991, empath: fface, emotion, gender recognition using holons, Advances in Neural Information Processing Systems 3, 564
  3. vanger, 1995, applications of synergetics in decoding facial expression of eemotion, Proc Third Int l Conf Automatic Face and Gesture Recognition, 24
  4. izard, 1971, The Face of Emotion
  5. johanstone, 1995, acoustic profiles in prototypical vocal expressions of emotions, Proc Int l Conf Phonetic Science, 4, 2
  6. kawakami, 1995, 3d emotion space for interactive communication, Proc Computer Science Conf, 471
  7. ekman, 1978, Facial Action Coding System (FACS) Manual
  8. fridlund, 1987, facial expressions of emotion: review literature 1970-1983, Nonverbal Behavior and Communication, 143
  9. ellis, 1986, process underlying face recognition, The Neuropsychology of Face Perception and Facial Expression, 1
  10. bruce, 1986, Recognizing Faces
  11. matsuno, 1993, recognition of facial expression with potential net, Proc Asian Conf Computer Vision, 504
  12. zhao, 1996, classifying facial emotions by backpropagation neural networks with fuzzy inputs, Proc Conf Neural Information Processing, 1, 454
  13. picard, 1999, offline and online recognition of emotion expression from physiological data, Emotion-Based Agent Architectures Workshop Notes Int l Conf Autonomous Agents, 135

Cited by

Sorted by citations

Related

Authors
Pantie, M., & Rothkrantz, L. J. M.
Affiliations
Delft University of Technology (NL)
ORCID
Rothkrantz, L.J.M.
Venue
IEEE Transactions on Pattern Analysis and Machine Intelligence
Publisher
Institute of Electrical and Electronics Engineers (IEEE)
Published
2000
Volume
22
Issue
12
Pages
1424-1445
ISSN
0162-8828 (ISSN-L)
DOI
10.1109/34.895976
OpenAlex
W2159017231
Semantic Scholar
087a6f471b5177014b08a342968901e1ef083ed1
Cited by
1,778
References
111
Citations by year
2026: 33 · 2025: 31 · 2024: 30 · 2023: 43 · 2022: 41 · 2021: 43 · 2020: 58 · 2019: 65 · 2018: 72 · 2017: 71 · 2016: 86 · 2015: 87 · 2014: 105 · 2013: 105 · 2012: 108
FWCI
23.88
Type
journal article
Language
en
Open access
Licence
ieeexplore.ieee.org/Xplorehelp/downloads/license-information/IEEE.html (vor)
Locations
  • IEEE Transactions on Pattern Analysis and Machine Intelligence (publishedVersion)
  • Research Repository (Delft University of Technology) (submittedVersion)
  • Data Archiving and Networked Services (DANS) (submittedVersion, other-oa)
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Face recognition and analysis
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Emotion and Mood Recognition
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Face and Expression Recognition
Keywords
facial expression analysis, face detection, facial expression recognition, emotion classification, human visual system, human-machine interaction, real-time systems

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-09.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-09.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review