xianyu

有计划实现类似python版的基于HMM的词性标注能力吗？

1

您好！ 目前cppjieba里的词性标注能力完全是基于词典和少量规则的，请问后续有计划实现类似python版posseg那样基于HMM的词性标注能力吗？

Some questions about building a tokenizer from scratch: vocab size can't decide actual vocab size and token order unstable.

11

Hi, I got some strange results when building a tokenizer from scratch followed by your offical demo.(https://huggingface.co/docs/tokenizers/python/latest/quicktour.html#build-a-tokenizer-from-scratch, and https://github.com/huggingface/tokenizers/blob/master/bindings/python/examples/train_bert_wordpiece.py) 1.The vocab size we set in training can't decide the actual...

More than one file was found with OS independent path 'LICENSE-EPL-1.0.txt' for android gradle

1

Execution failed for task ':app:mergeDebugJavaResource'. > A failure occurred while executing com.android.build.gradle.internal.tasks.Workers$ActionFacade > More than one file was found with OS independent path 'LICENSE-EPL-1.0.txt'.

关于 物品冷启动-利用物品的内容信息 中的遗留问题

5

@Magic-Bubble 您好，你在 物品冷启动-利用物品的内容信息 中提到的问题 “实验结果与书中的不符合(大多数指标明显偏低)，不知道是否是实现错误” 我尝试研究了几种可能的原因，发现在计算相似度时忽略余弦相似度的分母部分的话，结果会有大幅提升，接近书中的水平。 注释掉分母部分前后的结果： ``` for u in item_sim: for v in item_sim[u]: #可疑pos4 item_sim[u][v] /= math.sqrt(mo[u] * mo[v]) #余弦相似度 ``` Average Result (M=8, N=10, K=10): {'Precision':...