How can I JSON serialize an object from google's natural language API? (No __dict__ attribute)

I'm using the Google Natural Language API for a project tagging text with sentiment analysis. I want to store my NL results as JSON. If a direct HTTP request is made to Google then a JSON response is returned.

However when using the provided Python libraries an object is returned instead, and that object is not directly JSON serializable.

Here is a sample of my code:

import os
import sys
import oauth2client.client
from google.cloud.gapic.language.v1beta2 import enums, language_service_client
from google.cloud.proto.language.v1beta2 import language_service_pb2

class LanguageReader:
    # class that parses, stores and reports language data from text

    def __init__(self, content=None):

        try:
            # attempts to autheticate credentials from env variable
            oauth2client.client.GoogleCredentials.get_application_default()
        except oauth2client.client.ApplicationDefaultCredentialsError:
            print("=== ERROR: Google credentials could not be authenticated! ===")
            print("Current enviroment variable for this process is: {}".format(os.environ['GOOGLE_APPLICATION_CREDENTIALS']))
            print("Run:")
            print("   $ export GOOGLE_APPLICATION_CREDENTIALS=/YOUR_PATH_HERE/YOUR_JSON_KEY_HERE.json")
            print("to set the authentication credentials manually")
            sys.exit()

        self.language_client = language_service_client.LanguageServiceClient()
        self.document = language_service_pb2.Document()
        self.document.type = enums.Document.Type.PLAIN_TEXT
        self.encoding = enums.EncodingType.UTF32

        self.results = None

        if content is not None:
                self.read_content(content)

    def read_content(self, content):
        self.document.content = content
        self.language_client.analyze_sentiment(self.document, self.encoding)
        self.results = self.language_client.analyze_sentiment(self.document, self.encoding)

Now if you were to run:

sample_text="I love R&B music. Marvin Gaye is the best. 'What's Going On' is one of my favorite songs. It was so sad when Marvin Gaye died."
resp = LanguageReader(sample_text).results
print resp

You would get:

document_sentiment {
  magnitude: 2.40000009537
  score: 0.40000000596
}
language: "en"
sentences {
  text {
    content: "I love R&B music."
  }
  sentiment {
    magnitude: 0.800000011921
    score: 0.800000011921
  }
}
sentences {
  text {
    content: "Marvin Gaye is the best."
    begin_offset: 18
  }
  sentiment {
    magnitude: 0.800000011921
    score: 0.800000011921
  }
}
sentences {
  text {
    content: "\'What\'s Going On\' is one of my favorite songs."
    begin_offset: 43
  }
  sentiment {
    magnitude: 0.40000000596
    score: 0.40000000596
  }
}
sentences {
  text {
    content: "It was so sad when Marvin Gaye died."
    begin_offset: 90
  }
  sentiment {
    magnitude: 0.20000000298
    score: -0.20000000298
  }
}

Which is not JSON. It's an instance of the google.cloud.proto.language.v1beta2.language_service_pb2.AnalyzeSentimentResponse object. And it has no __dict__ attribute attribute so it is not serializable by using json.dumps().

How can I either specify that the response should be in JSON or serialize the object to JSON?

Edit: @Zach noted Google's protobuf Data Interchange Format. It seems the preferred option would be to use these protobuf.json_format methods:

from google.protobuf.json_format import MessageToDict, MessageToJson 

self.dict = MessageToDict(self.results)
self.json = MessageToJson(self.results)

From the docstring:

MessageToJson(message, including_default_value_fields=False, preserving_proto_field_name=False)
    Converts protobuf message to JSON format.

    Args:
      message: The protocol buffers message instance to serialize.
      including_default_value_fields: If True, singular primitive fields,
          repeated fields, and map fields will always be serialized.  If
          False, only serialize non-empty fields.  Singular message fields
          and oneof fields are not affected by this option.
      preserving_proto_field_name: If True, use the original proto field
          names as defined in the .proto file. If False, convert the field
          names to lowerCamelCase.

    Returns:
      A string containing the JSON formatted protocol buffer message.