StalemateBreaker: A Proactive Content-Introducing Approach to Automatic Human-Computer Conversation

Authors: Xiang Li, Lili Mou, Rui Yan, Ming Zhang

IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate our system on conversation logs from real-world users. Our approach outperforms several strong baselines as well as a state-of-the-practice system. ... We hired workers on a Chinese crowdsourcing platform to annotate all retrieved results with 1 Point (appropriate) or 0 Point (inappropriate). ... Table 1 shows the performance of our STALEMATEBREAKER system as well as a variety of baselines.
Researcher Affiliation Collaboration Xiang Li,1, Lili Mou,1,2 Rui Yan,3 Ming Zhang1 1School of EECS, Peking University, China {lixiang.eecs,mzhang cs}@pku.edu.cn 2Key Laboratory of High Confidence Software Technologies (Peking University), Ministry of Education, China doublepower.mou@gmail.com 3Natural Language Processing Department, Baidu Inc., China yanrui02@baidu.com
Pseudocode Yes Algorithm 1: Bi-Page Rank-HITS
Open Source Code No The paper does not include an explicit statement about releasing source code for the described methodology or a direct link to a code repository.
Open Datasets No The paper mentions collecting data from various platforms (Baidu Zhidao, Baidu Tieba, Douban forum, Sina Weibo) and using a knowledge graph mined from Baidu search logs, but it does not provide a direct URL, DOI, or specific repository name for accessing the collected dataset itself.
Dataset Splits No The paper describes the data collection and annotation process for evaluation but does not specify exact training, validation, or test dataset split percentages, sample counts, or refer to predefined splits.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU or CPU models, memory specifications) used to run the experiments.
Software Dependencies No The paper mentions using standard tools and models like Lucene, word2vec, and a learning-to-rank model, but it does not specify version numbers for any software dependencies.
Experiment Setup Yes In our Bi-Page Rank-HITS model, we have three main parameters, µ in the Page Rank phase, and x, y in the HITS phase. µ was set to 0.15 following Yan et al. [2012a] and not tuned in our experiment. For x and y, we tried different values with a granularity of 0.1.