Haystack 使用 Elasticsearch 建立索引时 修改为中文分词器

Haystack+Elasticsearch7 建立索引时默认的分词器为 snowball，比较适用于英文，

但显然对于中文分词来说并不友好，因此需将其更改为中文分词器

本项目结构如下图所示：

1 继承并重写 elasticsearch 搜索引擎

在 material 应用下新建名为 elasticsearch7_ik_backend.py 的文件，继承 Elasticsearch7SearchBackend（后端） 和 Elasticsearch7SearchEngine（搜索引擎） 并重写建立索引时的分词器设置

from haystack.backends.elasticsearch7_backend import Elasticsearch7SearchBackend, Elasticsearch7SearchEngine

DEFAULT_FIELD_MAPPING = {
    "type": "text",
    "analyzer": "ik_max_word",
    "search_analyzer": "ik_smart"
}


class Elasticsearc7IkSearchBackend(Elasticsearch7SearchBackend):
    def __init__(self, *args, **kwargs):
        self.DEFAULT_SETTINGS['settings']['analysis']['analyzer']['ik_analyzer'] = {
            "type": "custom",
            "tokenizer": "ik_max_word",
        }
        super(Elasticsearc7IkSearchBackend, self).__init__(*args, **kwargs)


class Elasticsearch7IkSearchEngine(Elasticsearch7SearchEngine):
    backend = Elasticsearc7IkSearchBackend

2 修改后端引擎 

在项目的配置文件 settings.py 中，将 haystack 原来的后端引擎 Elasticsearch7SearchEngine 修改为自己刚刚重新定义的 Elasticsearc7IkSearchBackend

HAYSTACK_CONNECTIONS = {
    'default': {
        # 项目名.文件名.搜索引擎名
        'ENGINE': 'material.elasticsearch7_ik_backend.Elasticsearch7IkSearchEngine',
        'URL': 'http://127.0.0.1:9200/', 
        'INDEX_NAME': 'template_v3.0',  # 索引库名称
    },
}

3 重建索引

使用修改分词器后的后端引擎，重新建立索引

python manage.py rebuild_index

补充

haystack 原先加载的是 ...\venv\Lib\site-packages\haystack\backends 文件夹下的 elasticsearch7_backend.py 文件，打开即可看到 elasticsearch7 引擎的默认配置

若用上述方法建立出来的索引字段仍使用 snowball 分词器，则将原先elasticsearch7_backend.py 文件中的 DEFAULT_FIELD_MAPPING 也修改为 ik 分词器（或许是因为版本问题）

DEFAULT_FIELD_MAPPING = {
    "type": "text",
    "analyzer": "ik_max_word",
    "search_analyzer": "ik_smart",
}

附：elasticsearch7_backend.py 文件原内容（需要被修改的部分）

DEFAULT_FIELD_MAPPING = {
    "type": "text",
    "analyzer": "ik_max_word",
    "search_analyzer": "ik_smart",
    # "analyzer": "snowball",
}

class Elasticsearch7SearchBackend(ElasticsearchSearchBackend):
    # Settings to add an n-gram & edge n-gram analyzer.
    DEFAULT_SETTINGS = {
        "settings": {
            "index": {
                "max_ngram_diff": 2,
            },
            "analysis": {
                "analyzer": {
                    "ngram_analyzer": {
                        "tokenizer": "standard",
                        "filter": [
                            "haystack_ngram",
                            "lowercase",
                        ],
                    },
                    "edgengram_analyzer": {
                        "tokenizer": "standard",
                        "filter": [
                            "haystack_edgengram",
                            "lowercase",
                        ],
                    },
                },
                "filter": {
                    "haystack_ngram": {
                        "type": "ngram",
                        "min_gram": 3,
                        "max_gram": 4,
                    },
                    "haystack_edgengram": {
                        "type": "edge_ngram",
                        "min_gram": 2,
                        "max_gram": 15,
                    },
                },
            },
        },
    }

    def __init__(self, connection_alias, **connection_options):
        super().__init__(connection_alias, **connection_options)
        self.content_field_name = None

    def build_search_kwargs(
            self,
            query_string,
            sort_by=None,
            start_offset=0,
            end_offset=None,
            fields="",
            highlight=False,
            facets=None,
            date_facets=None,
            query_facets=None,
            narrow_queries=None,
            spelling_query=None,
            within=None,
            dwithin=None,
            distance_point=None,
            models=None,
            limit_to_registered_models=None,
            result_class=None,
            **extra_kwargs
    ):
        index = haystack.connections[self.connection_alias].get_unified_index()
        content_field = index.document_field

        ......        

        return kwargs

    def _build_search_query_dwithin(self, dwithin):
        ......

    def _build_search_query_within(self, within):
        ......

    def _process_hits(self, raw_results):
        ......

    def _process_results(
            self,
            raw_results,
            highlight=False,
            result_class=None,
            distance_point=None,
            geo_sort=False,
    ):
        ......

    def _get_common_mapping(self):
        return {
            DJANGO_CT: {
                "type": "keyword",
            },
            DJANGO_ID: {
                "type": "keyword",
            },
        }

    def build_schema(self, fields):
        ......

        return (content_field_name, mapping)


class Elasticsearch7SearchQuery(ElasticsearchSearchQuery):
    def add_field_facet(self, field, **options):
        self.facets[field] = options.copy()


class Elasticsearch7SearchEngine(BaseEngine):
    backend = Elasticsearch7SearchBackend
    query = Elasticsearch7SearchQuery