使用机器学习套件提取实体 (iOS)

如需分析一段文本并提取其中的实体,请将文本直接传递给其 annotateText:completion: 方法,从而调用机器学习套件实体提取 API。此外,还可以传入包含其他配置选项(例如参考时间、时区或过滤条件)的可选 EntityExtractionParams 对象,以限制对部分实体类型的搜索。该 API 会返回 EntityAnnotation 对象列表,其中包含每个实体的相关信息。

实体提取基础检测器资源在应用运行时保持静态关联。这会给您的应用增加大约 10.7MB 的内存。

试试看

准备工作

  1. 在 Podfile 中添加以下机器学习套件库:

    pod 'GoogleMLKit/EntityExtraction', '3.2.0'
    
  2. 安装或更新项目的 Pod 之后,请使用 Xcode 项目的 .xcworkspace 来打开项目。Xcode 13.2.1 版或更高版本支持机器学习套件。

从文本中提取实体

如需从文本中提取实体,请先通过指定语言创建 EntityExtractorOptions 对象,然后使用该对象实例化 EntityExtractor

Swift

// Note: You can specify any of the 15 languages entity extraction supports here. 
let options = EntityExtractorOptions(modelIdentifier: 
                                    EntityExtractionModelIdentifier.english)
let entityExtractor = EntityExtractor.entityExtractor(options: options)

Objective-C

// Note: You can specify any of the 15 languages entity extraction supports here. 
MLKEntityExtractorOptions *options = 
    [[MLKEntityExtractorOptions alloc] 
        initWithModelIdentifier:MLKEntityExtractionModelIdentifierEnglish];

MLKEntityExtractor *entityExtractor = 
    [MLKEntityExtractor entityExtractorWithOptions:options];

接下来,确保将所需的语言模型下载到设备上:

Swift

entityExtractor.downloadModelIfNeeded(completion: {
  // If the error is nil, the download completed successfully.
})

Objective-C

[entityExtractor downloadModelIfNeededWithCompletion:^(NSError *_Nullable error) {
    // If the error is nil, the download completed successfully.
}];

下载模型后,将字符串和可选的 MLKEntityExtractionParams 传递给 annotate 方法。

Swift

// The EntityExtractionParams parameter is optional. Only instantiate and
// configure one if you need to customize one or more of its params.
var params = EntityExtractionParams()
// The params object contains the following properties which can be customized on
// each annotateText: call. Please see the class's documentation for a more
// detailed description of what each property represents.
params.referenceTime = Date();
params.referenceTimeZone = TimeZone(identifier: "GMT");
params.preferredLocale = Locale(identifier: "en-US");
params.typesFilter = Set([EntityType.address, EntityType.dateTime])

extractor.annotateText(
    text.string,
    params: params,
    completion: {
      result, error in
      // If the error is nil, the annotation completed successfully and any results 
      // will be contained in the `result` array.
    }
)

Objective-C

// The MLKEntityExtractionParams property is optional. Only instantiate and
// configure one if you need to customize one or more of its params.
MLKEntityExtractionParams *params = [[MLKEntityExtractionParams alloc] init];
// The params object contains the following properties which can be customized on
// each annotateText: call. Please see the class's documentation for a fuller 
// description of what each property represents.
params.referenceTime = [NSDate date];
params.referenceTimeZone = [NSTimeZone timeZoneWithAbbreviation:@"GMT"];
params.preferredLocale = [NSLocale localWithLocaleIdentifier:@"en-US"];
params.typesFilter = 
    [NSSet setWithObjects:MLKEntityExtractionEntityTypeAddress, 
                          MLKEntityExtractionEntityTypeDateTime, nil];

[extractor annotateText:text.string
             withParams:params
             completion:^(NSArray *_Nullable result, NSError *_Nullable error) {
  // If the error is nil, the annotation completed successfully and any results 
  // will be contained in the `result` array.
}

循环遍历注解结果,以检索有关所识别实体的信息。

Swift

// let annotations be the Array! returned from EntityExtractor
for annotation in annotations {
  let entities = annotation.entities
  for entity in entities {
    switch entity.entityType {
      case EntityType.dateTime:
        guard let dateTimeEntity = entity.dateTimeEntity else {
          print("This field should be populated.")
          return
        }
        print("Granularity: %d", dateTimeEntity.dateTimeGranularity)
        print("DateTime: %@", dateTimeEntity.dateTime)
      case EntityType.flightNumber:
        guard let flightNumberEntity = entity.flightNumberEntity else {
          print("This field should be populated.")
          return
        }
        print("Airline Code: %@", flightNumberEntity.airlineCode)
        print("Flight number: %@", flightNumberEntity.flightNumber)
      case EntityType.money:
        guard let moneyEntity = entity.moneyEntity else {
          print("This field should be populated.")
          return
        }
        print("Currency: %@", moneyEntity.integerPart)
        print("Integer Part: %d", moneyEntity.integerPart)
        print("Fractional Part: %d", moneyEntity.fractionalPart)
      // Add additional cases as needed.
      default:
        print("Entity: %@", entity);
    }
  }
}

Objective-C

NSArray *annotations; // Returned from EntityExtractor

for (MLKEntityAnnotation *annotation in annotations) {
            NSArray *entities = annotation.entities;
            NSLog(@"Range: [%d, %d)", (int)annotation.range.location, (int)(annotation.range.location + annotation.range.length));
            for (MLKEntity *entity in entities) {
              if ([entity.entityType isEqualToString:MLKEntityExtractionEntityTypeDateTime]) {
                MLKDateTimeEntity *dateTimeEntity = entity.dateTimeEntity;
                NSLog(@"Granularity: %d", (int)dateTimeEntity.dateTimeGranularity);
                NSLog(@"DateTime: %@", dateTimeEntity.dateTime);
                break;
              } else if ([entity.entityType isEqualToString:MLKEntityExtractionEntityTypeFlightNumber]) {
                MLKFlightNumberEntity *flightNumberEntity = entity.flightNumberEntity;
                NSLog(@"Airline Code: %@", flightNumberEntity.airlineCode);
                NSLog(@"Flight number: %@", flightNumberEntity.flightNumber);
                break;
              } else if ([entity.entityType isEqualToString:MLKEntityExtractionEntityTypeMoney]) {
                MLKMoneyEntity *moneyEntity = entity.moneyEntity;
                NSLog(@"Currency: %@", moneyEntity.unnormalizedCurrency);
                NSLog(@"Integer Part: %d", (int)moneyEntity.integerPart);
                NSLog(@"Fractional Part: %d", (int)moneyEntity.fractionalPart);
                break;
              } else {
                // Add additional cases as needed.
                NSLog(@"Entity: %@", entity);
              }
            }