Which Step Do I Take First? Troubleshooting with Bayesian Models
27184Authors
nan
Abstract
AbstractOnline discussion forums and community question-answering websites provide one of the primary avenues for online users to share information. In this paper, we propose text mining techniques which aid users navigate troubleshooting-oriented data such as questions asked on forums and their suggested solutions. We introduce Bayesian generative models of the troubleshooting data and apply them to two interrelated tasks: (a) predicting the complexity of the solutions (e.g., plugging a keyboard in the computer is easier compared to installing a special driver) and (b) presenting them in a ranked order from least to most complex. Experimental results show that our models are on par with human performance on these tasks, while outperforming baselines based on solution length or readability.
Keywords
Full Content
Which Step Do I Take First? Troubleshooting with Bayesian Models Annie Louis and Mirella Lapata School of Informatics, University of Edinburgh 10 Crichton Street, Edinburgh EH8 9AB {alouis,mlap}@inf.ed.ac.uk Abstract Online discussion forums and community question-answering websites provide one of the primary avenues for online users to share information. In this paper, we propose text mining techniques which aid users navigate troubleshooting-oriented data such as ques- tions asked on forums and their suggested so- lutions. We introduce Bayesian generative models of the troubleshooting data and apply them to two interrelated tasks: (a) predicting the complexity of the solutions (e.g., plugging a keyboard in the computer is easier compared to installing a special driver) and (b) present- ing them in a ranked order from least to most complex. Experimental results show that our models are on par with human performance on these tasks, while outperforming baselines based on solution length or readability. 1 Introduction Online forums and discussion boards have created novel ways for discovering, sharing, and distribut- ing information. Users typically post their ques- tions or problems and obtain possible solutions from other users. Through this simple mechanism of community-based question answering, it is possible to find answers to personal, open-ended, or highly specialized questions. However, navigating the in- formation available in web-archived data can be challenging given the lack of appropriate search and browsing facilities. Table 1 shows examples typical of the problems and proposed solutions found in troubleshooting- oriented online forums. The first problem concerns a shaky monitor and has three solutions with increas- ing degrees of complexity. Solution (1) is probably easiest to implement in terms of user time, effort, and expertise; solution (3) is most complex (i.e., the user should understand what signal timing is and The screen is shaking. 1. Move all objects that emit a magnetic field, such as a motor or transformer, away from the monitor. 2. Check if the specified voltage is applied. 3. Check if the signal timing of the computer system is within the specification of the monitor. “Illegal Operation has Occurred” error message is displayed. 1. Software being used is not Microsoft-certified for your version of Windows. Verify that the software is certified by Microsoft for your version of Windows (see program packaging for this information). 2. Configuration files are corrupt. If possible, save all data, close all programs, and restart the computer. Table 1: Example problems and solutions taken from on- line troubleshooting oriented forums. then try to establish whether it is within the specifi- cation of the monitor), whereas solution (2) is some- where in between. In most cases, the solutions are not organized in any particular fashion, neither in terms of content nor complexity. In this paper, we present models to automatically predict the complexity of troubleshooting solutions, which we argue could improve user experience, and potentially help solve the problem faster (e.g., by prioritizing easier solutions). Automatically struc- turing solutions according to complexity could also facilitate search through large archives of solutions or serve as a …
Content truncated to 500 words
Labeled Text
AbstractOnline discussion forums and community question-answering websites provide one of the primary avenues for online users to share information. In this paper, we propose text mining techniques which aid users navigate troubleshooting-oriented data such as questions asked on forums and their suggested solutions. We introduce Bayesian generative models of the troubleshooting data and apply them to two interrelated tasks: (a) predicting the complexity of the solutions (e.g., plugging a keyboard in the computer is easier compared to installing a special driver) and (b) presenting them in a ranked order from least to most complex. Experimental results show that our models are on par with human performance on these tasks, while outperforming baselines based on solution length or readability.
Publication Details
- Paper ID:
- 27184
- Year:
- 2015
- Publication:
- TIPSTER
- Publication Type:
- conf
- DOI:
- nan