Introduction
In the mid-2000s, Peer-to-Peer (P2P) systems like Maze emerged as massive repositories of shared data. However, these decentralized networks suffered from semantic heterogeneity. Without naming standards, users frequently labeled identical files differently based on their native language. For instance, a single movie file might be named "movie" in English, "电影" in Chinese characters, or "dianying" in Pinyin. Because standard P2P search engines typically return results matching only the specific language of a query, these naming conflicts fragment the network, reducing resource availability and slowing download speeds.
Methodology
To bridge this "language gap," this paper proposes a computer-aided framework to construct a dictionary of file correspondences. The system relies on the mathematical properties of equivalence—reflexivity, symmetry, and transitivity—to link synonyms across English, Chinese, and Pinyin.
The architecture consists of three primary components:
- Chinese-English Translator: Utilizes online bilingual dictionaries and domain-specific thesauruses. To resolve ambiguities (e.g., distinguishing "movie" from "cinema"), the system incorporates human intervention to select the most relevant terms.
- Pinyin Translator: Automatically maps Chinese names to Pinyin using linguistic tools, requiring minimal manual oversight.
- Matching Module: Compares candidate names against structural characteristics (such as file extensions) of files retrieved from different nodes to verify resource identity.
By establishing that "movie" corresponds to "电影" and "电影" to "dianying," the system uses transitivity to automatically link "movie" to "dianying," allowing the dictionary to scale efficiently.
Experimental Setup
The approach was evaluated using data from the Maze system, a hybrid P2P community developed at Peking University. At the time of study, Maze hosted over 140 million files shared by 210,000 users. The researchers analyzed file name distributions across categories like Video and Software and tested an experimental prototype using a random sample of 10,000 files.
Key Results
The integrated "ChnEngPY" (Chinese-English-Pinyin) dictionary demonstrated a dramatic improvement in search effectiveness compared to single-language queries.
As shown in the experimental results, searching with only English names yielded a recall of approximately 42.7%. By utilizing the multi-language dictionary, recall surged to 83.6%. Crucially, this broader search did not significantly compromise accuracy; precision remained high at 90.1%, compared to the 91-92% range for single-language searches. This increased recall allows users to identify more "mirror" copies of a file, facilitating simultaneous downloads from multiple sources and significantly improving download efficiency.
Limitations and Future Work
While effective, the current heuristic approach relies heavily on file names and basic structural features, which can occasionally be misled by unrelated files with similar names. Additionally, the system still requires domain experts for dictionary verification. Future research aims to incorporate content-based analysis—examining the actual data within files—to further automate the process and enhance the accuracy of the correspondence dictionary.
Summary updated