Automated Inference of Socio-Cultural Information from Natural Language Conversations

We discuss a methodology for extracting socio-cultural information from transcripts of natural language conversations. The methodology is applicable to a wide variety of languages. We use Russian and Tamil for illustration. The extracted socio-cultural information pertains to the nature of the relationship between the participants in the interaction. We concentrate on the information implicit in the use of terms that refer to people (pronouns, terms of address etc.). We have constructed an AnsProlog theory of the use of these language indicators in Russian and also in Tamil. It is this theory that looks at the usages of these indicators in the conversation in question and produces information about the relationships of the participants in the conversation.