baidu / baidu/lac

歌名实体识别失败

Open
#67 1 comment 0 reactions 0 assignees View on GitHub
Dominant language
C++
Stars
4k
Forks
592
PR merge metrics
No merged PRs in 30d

Description

尝试使用lac对歌名进行实体识别,按照词性表,期望歌名标记出nw标签,但是观察到:
1. 歌名往往被拆分
2. 几乎没有歌名被正确标注为nw

原因主要在于训练数据没有针对歌名实体进行适配?

例如:
```python
来/v 首生/v 偏字/n
来/v 一首/m 再会/v 无缘/v 的/u 情人/n
给/v 我/r 点/v 一首/m 凡人歌/n
请/v 唱一首/v 呼伦贝尔大草原/LOC
来/v 一首/m 像/v 我/r 这样的人/n
唱/v 首/q 赵雷/PER 的/u 成都/LOC
来/v 一首/m 起风/v 了/xc
听/v /w dj/nz /w 版/n 的/u 这首歌/n
下/f 一首歌/n
来/v 一首/m 错过了/v 缘分/n 错过了/v 你/r
来/v 一首/m
来/v 一首/m 出山/v
播放/v 一首/m 丢了/v 幸福/a 的/u 猪/n
唱一首/v 大衣哥/PER 的/u 歌曲/n
播放/v 一首/m 一曲/n 相思/v
来/v 一首/m 爱的世界/n 只有你/v
来/v 一首/m 缘分/n
点播/v 一首/m 歌曲/n
感情戏/n 这/r 首/q 歌曲/n
播放/v 一首/m 不在/v 犹豫/a /w dj/nz /w 版/n 的/u
唱一首/v
放/v 一首/m 刀郎/PER 的/u 歌/n
来/v 一首/m 离/v 人/n 愁/v
```

Contributor guide

No contributing guide indexed for this repository

Research direction

Start by reproducing the listed Chinese song-title examples with LAC and compare segmentation and entity tags against the expected nw label. Investigate whether the training data covers song-title entities, then define completion as song names being kept together and correctly labeled nw across representative examples.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.