A Survey on Email Summarisation Techniques and its Challenges

Abstract

We have reviewed various papers such as “Summarizing Emails with Conversational Cohesion and Subjectivity†and “Task-focused Summarization of Emailâ€. In paper “Summarizing Emails with Conversational Cohesion and Subjectivity†[12] they have studied how to summarize email conversations based on the conversational cohesion and the subjective opinions. They first create a sentence quotation graph to represent the conversation structure on the sentence level. They adopted three cohesion metrics, clue words, semantic similarity and cosine similarity, to measure the weight of the edges. The Generalized Clue Word Summarizer and Page-Rank are applied to this graph to produce summaries. Moreover, they study how to include subjective opinions to help identify important sentences for summarization. The empirical evaluation shows the following two discoveries: (1) the basic CWS (based on clue words) obtains a higher accuracy and a better runtime performance than the other cohesion measures. It also has a significant higher accuracy than the Page-Rank algorithm. (2) By integrating clue words and subjective words (phrases), the accuracy of CWS is improved significantly. This reveals an interesting phenomenon and will be further studied. Whereas in other paper “Task-focused Summarization of Email†[11]. They have presented aspects of Smart Mail, which provides a task-oriented summary of email messages. This summary is produced by identifying the task-related sentences in the message and then reformulating each task related sentence as a brief (usually imperative) summation of the task. The set of tasks extracted and reformulated from a given email message is thus a task-focused summary of that message. The ability to reformulate tasks is in principle separate from the identification of tasks. In their planned usability study they will distribute variants of the prototype to determine the effect of reformulation.