A Ph.D. proposal talk by Snigdha Chaturvedi, who is expected to graduate in June 2016. The presentation was terrific covering relationship prediction in movies, novels and MOOCs.
Time:  06.30.2015 15:00 to 16:30
Location:  AVW 4424

In this thesis we propose methods to model inter-personal relationships in text. Due to their inherent social nature, people continuously interact and form relationships with each other. Understanding these relationships is essential to understanding and explaining people’s desires, goals, actions and expected behaviors. Apart from applications related to general natural language understanding, modeling inter-personal relationships also finds application in many real-world domains such as social networks, discussion forums etc. In this proposal we provide methods to model inter-personal relationships in natural language text with a focus on narratives. We demonstrate that such a task can benefit from using models that are capable of incorporating not just linguistic cues but also the contexts in which these cues appear. We consider two types of narratives: movies and novels, and propose structured models to address the task of modeling the nature of relationships between any two given characters from the narrative. We attempt to jointly infer the nature of relationships between all characters in the narrative and demonstrate how the task of identifying relationship between two characters can benefit by including information about their relationships with other characters in the movie. We next formulate the relationship-modeling problem as a structured prediction task to acknowledge the evolving nature of human relationships and demonstrate the need to model history of relationships between characters while modeling their current relationship. We then propose to jointly address the relationship-modeling task in the two domains mentioned above to better utilize their commonalities while simultaneously teasing apart their idiosyncrasies. Lastly, we demonstrate a practical application of this task. We analyze contents of online educational discussion forums to automatically suggest threads to the instructors that require their intervention. By suggesting avenues for instructor-student interaction, we alleviate the need for the instructor to manually go over all threads of the forum and also help the students who have no way of interacting with the instructor. We propose to incorporate thread structure into our approach by using latent variables that abstractly represent contents of individual posts and model the flow of information in the thread. Examining Committee:

Committee Chair: Dr. Hal Daume III (Prof. Daume’s homepage is extremely interesting)

Dept’s Representative Dr. Hector Corrada Bravo

Committee Member(s): Dr. Chris Dyer (CMU, LTI) Dr. Philip Resnik

Structured Approaches to Exploring Inter-Personal Relationships in Natural Language Text


People interact via language. Now we have digitally available texts, especially narratives are full of social interactions: Movies novels, plays, blogs, news articles and etc.

Why is relationship modeling important?

Because Natural Language Understanding understands peoples’ behavior and goals

  • Social networks – recommend ‘friend’, personalize news feeds
  • Smarter emails – recommend recipients
  • Community Question answering – recommend questions

Goal: present methods to model inter-personal relationships in NLP with a focus on narratives; use structured models and demonstrate the importance

Relationship Identification in Movies

  • Domain: Movie Structures
    • Existing character-centric approaches model roles
    • Sometimes narratives are more about relationships
  • Problem formalism
    • Given character 1 and character 2, predict cooperative / non-cooperative relationship
  • Assumption
    • Real-life relationships have multiple facets
    • We consider binary relationships with focus on teasing point
  • Supervised approach
    • Features
      • Linguistic
      • Semantic
    • Contextual / Structural: Consider relationships of the two characters with others in the movie
  • Relationship
    • Relationship are not static but evolve with the progress of the narrative
    • Clique (cooperative x3), love triangle (coop x2), common enemy (coop x1), Mexican standoff (non-cooperative x3) // this is an interesting summary
    • Love triangle and Mexican standoff are balanced / stable.
  • Logistic Regression (unstructured) is compared with structured model

Evolving Relationships in Novels

  • Relationship sequence in novels (4th dimension)
    • Partially-supervised
    • Historical context: 2nd order Markov assumption
  • Example
    • Ron and Hermione were cooperative in the Deathly Hallow but non-cooperative in the Prisoner of Azkaban…
  • Modeling
    • Relationship Modeling: Each character pair annotated with relationship sequence.
    • Change Detection: Each character pair annotated with whether relationship changed at least once. Summary based on summary.

Joint Relationship Models Across Domains

  • Similarity
    • Movies and novels both describe relationships between multiple characters and used similar features
  • Difference
    • Movies are dramatic events within a short period
    • Novels has greater number of characters with more subtleties of relationships
  • Proposed work
    • Joint modeling for muliple narrative domains
  • score = \alpha w_g \cdot \phi_g (text, r) + (1 – \alpha) w_d \cdot \phi_d (text, r)

Application: MOOC

  • Input: given a thread (t), a chain of posts (p_1, p_2, …, p_n), each post belongs to a conceptual categories (latent): 1, 2, 3, …, K
  • Output: Intervention or not.
  • Intervention is initiation of a relationship
    • Students post seeking attention: request for relationship
    • Instructor replies
  • Latent Chain Markov Model (LCMM)
    • Setting: supervised (reply state given)
    • Prediction: uses final post
    • Score: f_w(t, p) = \max_h [w \cdot \phi(p, r, h, t)]
    • Data
      • Threads from forums for two MOOCs: GHC, WCR
      • Each thread annotated with binary reply state

Dissertation provides contribution to knowledge.

– Prof. Philip Resnik