Update README.md
Browse files
README.md
CHANGED
|
@@ -44,7 +44,19 @@ Utilizing the re-ranking model (e.g., [bge-reranker](https://github.com/FlagOpen
|
|
| 44 |
- Sparse retrieval (lexical matching): a vector of size equal to the vocabulary, with the majority of positions set to zero, calculating a weight only for tokens present in the text. e.g., BM25, [unicoil](https://arxiv.org/pdf/2106.14807.pdf), and [splade](https://arxiv.org/abs/2107.05720)
|
| 45 |
- Multi-vector retrieval: use multiple vectors to represent a text, e.g., [ColBERT](https://arxiv.org/abs/2004.12832).
|
| 46 |
|
| 47 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 48 |
|
| 49 |
For embedding retrieval, you can employ the BGE-M3 model using the same approach as BGE.
|
| 50 |
The only difference is that the BGE-M3 model no longer requires adding instructions to the queries.
|
|
@@ -52,7 +64,7 @@ For sparse retrieval methods, most open-source libraries currently do not suppor
|
|
| 52 |
Contributions from the community are welcome.
|
| 53 |
|
| 54 |
|
| 55 |
-
**
|
| 56 |
|
| 57 |
You can follow the common in this [example](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune)
|
| 58 |
to fine-tune the dense embedding.
|
|
|
|
| 44 |
- Sparse retrieval (lexical matching): a vector of size equal to the vocabulary, with the majority of positions set to zero, calculating a weight only for tokens present in the text. e.g., BM25, [unicoil](https://arxiv.org/pdf/2106.14807.pdf), and [splade](https://arxiv.org/abs/2107.05720)
|
| 45 |
- Multi-vector retrieval: use multiple vectors to represent a text, e.g., [ColBERT](https://arxiv.org/abs/2004.12832).
|
| 46 |
|
| 47 |
+
|
| 48 |
+
**2. Comparison with BGE-v1.5 and other monolingual models**
|
| 49 |
+
|
| 50 |
+
BGE-M3 is a multilingual model, and its ability in monolingual embedding retrieval may not surpass models specifically designed for single languages.
|
| 51 |
+
However, we still recommend trying BGE-M3 because of its versatility (support for multiple languages and long texts).
|
| 52 |
+
Moreover, it can simultaneously generate multiple representations, and using them together can enhance accuracy and generalization,
|
| 53 |
+
unlike most existing models that can only perform dense retrieval.
|
| 54 |
+
|
| 55 |
+
In the open-source community, there are many excellent models (e.g., jina-embedding, colbert, e5, etc),
|
| 56 |
+
and users can choose a model that suits their specific needs based on practical considerations,
|
| 57 |
+
such as whether to require multilingual or cross-language support, and whether to process long texts.
|
| 58 |
+
|
| 59 |
+
**3. How to use BGE-M3 in other projects?**
|
| 60 |
|
| 61 |
For embedding retrieval, you can employ the BGE-M3 model using the same approach as BGE.
|
| 62 |
The only difference is that the BGE-M3 model no longer requires adding instructions to the queries.
|
|
|
|
| 64 |
Contributions from the community are welcome.
|
| 65 |
|
| 66 |
|
| 67 |
+
**4. How to fine-tune bge-M3 model?**
|
| 68 |
|
| 69 |
You can follow the common in this [example](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune)
|
| 70 |
to fine-tune the dense embedding.
|