machine-learning - 我可以在 CRF 模型中使用数值特征吗

Question

在 CRF 模型中添加数值特征是否可能/很好？例如序列中的位置。

我正在使用CRFsuite 。似乎所有功能都会转换为字符串，例如'pos=0'、'pos=1'，这样就失去了欧氏距离的含义。

或者我应该使用它们来训练另一个模型，例如svm，然后与 cRF 模型集成？

Answer 1

我发现 CRFsuite 确实可以处理数字特征，至少根据this documentation :

• {“string_key”: float_weight, ...} dict where keys are observed features and values are their weights;
• {“string_key”: bool, ...} dict; True is converted to 1.0 weight, False - to 0.0;
• {“string_key”: “string_value”, ...} dict; that’s the same as {“string_key=string_value”: 1.0, ...}
• [“string_key1”, “string_key2”, ...] list; that’s the same as {“string_key1”: 1.0, “string_key2”: 1.0, ...}
• {“string_prefix”: {...}} dicts: nested dict is processed and “string_prefix” s prepended to each key.
• {“string_prefix”: [...]} dicts: nested list is processed and “string_prefix” s prepended to each key.
• {“string_prefix”: set([...])} dicts: nested list is processed and “string_prefix” s prepended to each key.

只要:

1. 我保持输入格式正确；
2. 我使用 float 与 float 字符串；
3. 我将其标准化。