Automated story capture from conversational speech

Abstract

While storytelling has long been recognized as an important part of effective knowledge management in organizations, knowledge management technologies have generally not distinguished between stories and other types of discourse. In this paper we describe a new type of technological support for storytelling that involves automatically capturing the stories that people tell to each other in conversations. We describe our first attempt at constructing an automated story extraction system using statistical text classification and a simple voting scheme. We evaluate the performance of this system and demonstrate that useful levels of precision and recall can be obtained when analyzing transcripts of interviews, but that performance on speech recognition data is not above what can be expected by chance. This paper establishes the level of performance that can be obtained using a straightforward approach to story extraction, and outlines ways in which future systems can improve on these results and enable a wide range of knowledge socialization applications. Categories and Subject Descriptors I.2.7 Natural Language Processing General TermsAlgorithms. KeywordsStorytelling, Knowledge Management. KNOWLEDGE SOCIALIZATIONMuch of the knowledge that is shared among members of communities and organizations is exhibited only in the telling of stories in spoken conversations. While support for storytelling in organizations has been long recognized as important to effective knowledge management [2][12], few attempts have been made to specifically support storytelling through technology. Instead of creating specific technologies for automatically capturing, analyzing, and routing stories that are naturally told in conversations, knowledge management technology development has targeted the more general problem of supporting computer-mediated communication, without much regard to the genre of the content [9]. In not distinguishing between storytelling and other types of human-human communication, today's knowledge management technologies fail to exploit the value of stories in packaging and transmitting tacit knowledge [13], understanding organizational change [10], and driving the development of professional training applications [5]. Furthermore, knowledge management technologies have had difficulty breaking out of the mold of traditional networked groupware applications, which limits their applicability to the fraction of people who spend their days working at computer terminals. A different vision for knowledge management technology is one that is specifically targeted to the capture and use of the stories told in communities and organizations in the context of normal, spoken conversations. The role of technology would be to support the capture of stories from spoken conversations, perform a task-directed analysis of its content, and present stories or analysis of stories to people in service of their organizational tasks. As a hypothetical example, consider the utility of story management technology for militaries with...

A Semantic Triplet Based Story Classifier

“…Gordon investigated the problem of detecting stories in conversational speech [7] and weblogs [8] and [9]. In [7], the authors train a Naive Bayes classifier to categorize the transcribed text of a speech into story and non-story categories.…”

“…In [7], the authors train a Naive Bayes classifier to categorize the transcribed text of a speech into story and non-story categories. Using word-level unigram and bigram frequency counts as feature vectors, they reported results for the classification of a speech as a story with 53.0% precision, 62.9% recall and 0.575 F-measure.…”

A Data-Driven Case-Based Reasoning Approach to Interactive Storytelling

“…To automatically identify stories in the ICWSM corpus we followed a similar approach to Gordon & Ganesan [8], which takes a binary classification approach to the problem. However, in our work, entire blog posts are labeled as a positive or negative instance of a personal story, as opposed to extracting smaller segments from the text.…”

Automated story capture from internet weblogs

“…There were an estimated 70 million weblogs in March 2007 (http://www.technorati.com), each with a series of entries containing the thoughts and ramblings of some computer user. We analyzed the entries of a random sample of 100 weblogs (approximately 12,000 sentences or 200,000 words) and found that 17% of weblog text consisted of stories, using the annotations guidelines of previous research [1]. By extrapolation, we can estimate that there are 23.8 billion words of story text available on the web for use in knowledge management and training applications.…”

“…Gordon & Ganesan [1] first demonstrated the feasibility of automatically extracting stories from text using statistical natural language processing techniques. The aim of their work was to develop technologies for identifying stories in conversational speech data, e.g.…”