1 / 26

Action Verb Processing in Indian Languages:The Role of Body Part Knowledge

Action Verb Processing in Indian Languages:The Role of Body Part Knowledge. Vasanta D 1 , Bapi Raju S 2 ., Suvarna A 3 ., Patel J 4 ., Viswanatha Naidu Y 5 ., Sireesha J. 6 and Nigam R. 7 1, 5 Department of Linguistics, Osmania University, Hyderabad.

romeo
Télécharger la présentation

Action Verb Processing in Indian Languages:The Role of Body Part Knowledge

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Action Verb Processing in Indian Languages:The Role of Body Part Knowledge Vasanta D1, Bapi Raju S2., Suvarna A3., Patel J4., Viswanatha Naidu Y5., Sireesha J.6 and Nigam R.7 1, 5 Department of Linguistics, Osmania University, Hyderabad. 2, 4 Dept. of Computer and Information Sciences, University of Hyderabad; 3, 6, 7 Department of Neurology, Nizam’s Institute for Medical Sciences, Hyderabad

  2. Action Verbs • Children learn verbs referring to actions performed by specific body parts such as for instance: • Legs (e.g. kick), • Hands (e.g. wave) or • Mouth (e.g. chew) relatively early either because they perform those actions or observe others performing them (Huttenlocher, J., Smiley, P., and Charney, R. – 1983 ) .

  3. Meaning of Action verbs • In a series of four different experiments Bergen et al. (2009) demonstrated that accessing meaning of action verbs such as ‘smile’, ‘kick’ and ‘punch’ require language users to activate modality specific cognitive representations responsible for performing and perceiving those actions. • In this study involving both native and non-native English speakers it was noted that learners progressively develop the ability to activate neural circuitry that supports grounded modality specific representations.

  4. Neural substrate for Action words • Action words are thought of as abstract semantic links between language elements and motor programmes (Pulvermuller, F. 2005 ) • Premotor and parietal areas are neurally integrated not only to control action but also to help construct an integrated representation of actions, objects acted upon, and the locations towards which the actions are directed (Feldman, J. and Narayanan, S. 2004)

  5. Structure vs. function systems for action More recent research (Buxbaum and Kalenine 2010) suggests that two separate systems exist that are differently specialized for stored functional knowledge (involving dorsal pathways) versus online calculations of structural information (involving ventral pathways) –Jigar see if you can include Fig.4, P. 208 from this article While functional representations may induce long lasting interference as is typical of semantic memory, whereas structural representations are brief and are likely to decay rapidly. Buxbaum, L.J. and Kalenine, S. 2010. Action knowledge, visuomotor activation, and embodiment in the two action systems)

  6. Action naming vs. Object naming • Production of verbs required for action naming is said to be better preserved in Alzheimer’s patients than production of nouns required for object naming (Druks et al, 2006). • Others reported that action naming impairment is not related to damage to one specific cortical region, instead it is associated with conditions that affect the frontoparietal-subcortical circuits involved in action knowledge and action representation (Cotelli. et al 2006.).

  7. Action and body part segmentation • Intentional action imposes a functional organization to body parts. • HAND is a crucial concept required to explain meanings of many action words (e.g. slap, stroke, tear etc), it must be universally lexicalized. English speakers associate these verbs with hand (Maouene et al. 2008). • However, slap could be achieved through PALM and tear and stroke through FINGERS. There is a need therefore to investigate verb body part associations across different language users to enhance our understanding about action representations.

  8. Variation in body part naming Many languages possess the term ‘BODY’, but there are some languages that do not have a term for it. Most languages distinguish among HEAD, TRUNK, ARMS, LEGS Many languages lack a distinct word for TRUNK (TORSO), whereas there are languages in which TRUNK includes CHEST, BELLY, and BUTTOCKS. English does not include BUTTOCKS as part of TRUNK Joints constitute delimitation points for the extension of body part terms. (see Majid, A. 2010.Words for parts of the body. In B.C. Malt and P. Wolff (Eds.), Words in the mind. OUP. For a detailed discussion Jigar, a figure will help to illustrate different geons - partonomy

  9. Verb body part associations among English native speakers • Maouene, Hidaka and Smith (2008) collected in a free association task, body parts associated with 101 early learned action verbs from 50 English speaking college students. • They reported obtaining sixty one unique body part words from all the participants for the target verbs • Consistently high agreement was noted for body parts associated with hand / arm, leg and mouth in that order • Words with high association strength were reported to be highly concrete verbs that clearly labeled actions performed by specific body parts.

  10. Action verb- body part association in multilingual adults: Present study • Do people growing up in bi or multilingual language environments organize their body part lexicons differently from monolinguals? • What is the nature of verb-body part associations in multilingual adults whose language use histories are different? • The present study addressed these questions using data from Hindi, Telugu, Dakkhini Urdu (L1) and English (L2) speaking normal adults and compared the results to English (L1) speakers of Maouene et al (2008) study as well as preliminary results based on 12 patients diagnosed to have Alzheimer’s disease.

  11. The Experiment

  12. The Experiment - Data Collection • One hundred target action verbs were randomized, and four different sets of data sheets were generated. • The verbs were printed in the orthography associated with each language in the first column. • The second column on the data sheet had the heading ‘body-part’. • The participants were asked to read the target verb, and write against each verb, the name of the body part that comes to their mind as soon as possible.

  13. Data Analysis • Data were analyzed to find out • The number of unique body part labels used in each language, • The number of body parts required in order to obtain 80% association agreement • Body part-body regions across languages • Common verbs across languages on which 80% association agreement is present • Comparison with English (L1) data and with results based on AD patients

  14. Unique body parts: Normal adults • Unique body Parts Labels: • Noticeable cross-linguistic differences were observed in the number of unique body parts reported language-wise and as compared to results of Maouene et al 2008 study.

  15. Coverage of 85% associations • The number of body parts that accounted for 85 % o of the associations in each language varied slightly. For Hindi only 5 body parts, for Telugu, Urdu and English, 6, 6 and 7 body parts respectively were required to cover 85% of all the associations. • For English native speakers of Maouene et al (2008) study, the corresponding number was 15. It appears that Indian language users are using more general body part terms compared to native English speakers.

  16. Verbs receiving 80% agreement • Most of these verbs are associated with hand, leg and mouth in that order for all the four language groups.

  17. Verbs receiving 80% agreement • The number of Verbs in each language that met minimum 80 % of agreement is shown in the bar chart and last column of the table.

  18. Most important body parts • Most of the 100 verbs are associated with THREE major body parts. Namely, Hand Leg, and Mouth, the first three colours from the bottom of the chart.

  19. Common verbs with good agreement • 12 of the 100 target verbs received consistently high agreement in all the four languages with reference to the body parts, Hand, Leg and Eye. • These verbs are visually salient as evident by the visual valency (VV) ratings.

  20. Visual Valency of the verbs • Visual Valency refers to the capacity of a verb to take a specific number of visual arguments (see Ma et al 2006. Visual semantics and ontology of eventive verbs). • The 12 verbs listed in the previous slide need two or more visual arguments to include semantic components of agent, object and instrument/location connected to the action depicted by the verb in a visual domain. • Note that all the hand-related verbs have 3 or more visual valencies (For instance a verb like ‘ko:yu/ka:Tna:’ ‘cut’ has three visual arguments, namely agent, object and instrument whereas, e:Dcu / ro:na ‘to cry’ has only one vv).

  21. Body region vs. body part • Using cluster analysis, Maouene et al. 2008 had identified five body regions each of which received different labels for specific body parts belonging to that region. • Similar clustering of the body regions done in our study revealed slight differences. • The Indian language users did not use parts such as knuckle, but specified wrist, nail and biceps as part of HAND; did not use thigh and toe as belonging to LEG.

  22. Body region-body part knowledge • Since body part categorization seems to be somewhat different in native speakers of English (Maouene et al 2008 study) from that of users of English in India (where it is a second or a third language), we have looked at the association agreement between body region and body part for nine common verbs used in both the studies.

  23. Body region vs body part (English data)Jigar, please redo this table- not clear

  24. Summary of results on normal adults • The body part lexicons associated with the three Indian languages seem to differ somewhat. Dakkhini Urdu and L2 English have more elaborate lexicons than Telugu and Hinid, but not as extensive as L1 English. • Indian Language speakers on the whole are using more super-ordinate general body parts like, HAND, LEG and MOUTH compared to L1 English speakers • Actions performed by HAND in which vision plays a role seem to have cross-linguistic agreement with the body parts associated with those actions.

  25. Preliminary results: AD patients • The 12 AD patients in our study (average age: 69.16) • Average education: 13.41 • exhibited knowledge of body part names that are not very different from those based on normal adults • However, they required 30 verbs, all involving actions of HAND on which there was 80% agreement suggesting that their knowledge of HAND and its involvement in action verbs is more robust (did not breakdown) compared to other body parts.

  26. Conclusions • Categories associated with our bodies and their actions are language specific. • Super-ordinate body parts (e.g. HAND) seem to relate to functions in general, whereas, parts at subordinate level (PALM, FINGERS) have ambiguous status in that they refer to appearance and function at the same time. Action requires unified representation of the body that puts parts back together. • Brain damage associated with AD affects unified representation of the body, but seems to preserve action knowledge involving major body parts such as HANDS.

More Related