Question

我正在使用LUIS Authoring API将培训数据发送到LUIS应用程序。我看到两个几乎相同的训练话语一致的索引错误：

POST数据

[
    {
        "text": "what are the main drivers for Aon",
        "intentName": "QI.StockOverview",
        "entityLabels": [
            {
                "entityName": "company",
                "startCharIndex": 30,
                "endCharIndex": 32
            }
        ]
    },
    {
        "text": "what are the main drivers for Concho Resources",
        "intentName": "QI.StockOverview",
        "entityLabels": [
            {
                "entityName": "company",
                "startCharIndex": 30,
                "endCharIndex": 45
            }
        ]
    }
]

我故意将endCharIndex减少一个，假设它意味着＆＃34;最多但不包括最后一个字符＆＃34;在某些语言中很常见。但是，我仍然得到这样的答复：

[
    {
        "value": {
            "ExampleId": 66579136,
            "UtteranceText": "what are the main drivers for aon"
        },
        "hasError": false
    },
    {
        "value": null,
        "hasError": true,
        "error": {
            "code": "FAILED",
            "message": "Example: \"what are the main drivers for Concho Resources\". Error: Index and length must refer to a location within the string.\r\nParameter name: length"
        }
    }
]

当切出字符串中不存在的索引时，该消息似乎是标准的C＃错误。让我感到困惑的是，其中一个话语起作用而另一个话语起作用。

文档here说＆＃34;两个索引都是从零开始计算的＆＃34;，我相信我做得不错。

任何提示？感谢

编辑：

我正在使用与this demo大致相同的Python脚本。我只进行了一次更改，即90个批次的POST，因为我有900个培训话语，每个请求的限制似乎是100.相关的方法是：

def add_utterances(self, filename=UTTERANCE_FILE):
    with open(filename, encoding=self.UTF8) as utterance:
        data = json.load(utterance)

    chunk_size = 90
    print(f'Adding {len(data)} utterances in chunks of {chunk_size}')
    for i, chunk in enumerate(chunks(data, chunk_size)):
        print('chunk', i)
        self.call(self.EXAMPLES, self.POST, json.dumps(chunk))
        print(self.result)

HTTP请求是

POST westus.api.cognitive.microsoft.com/luis/api/v2.0/apps/{app_id}/versions/{app_version}/examples

将JSON列表作为请求正文

Microsoft LUIS创作API

0 个答案: